On 14/07/2022 6:13 am, Richard Purdie wrote:
On Tue, 2022-07-12 at 13:23 +0800, Phil Reid wrote:
On 8/07/2022 9:05 pm, Richard Purdie wrote:
On Tue, 2022-07-05 at 21:01 -0700, [email protected] wrote:
Hello there,
It isn't due to EOVERFLOW but make_file_type function in libstdc++-
v3\src\filesystem\ops-common.h always returns file_type::unknown.
It's because _GLIBCXX_HAVE_S_ISREG isn't defined.
I found HAVE_S_ISREG at libstdc++-v3/config.h isn't set properly.
It's caused by a configure failure below to tell if S_ISREG exists on
the target.
gcc/arm-emit-linux-gnueabi/11.3.0/ld: error: tmp-
glibc/work/cortexa9t2hf-neon-emit-linux-gnueabi/gcc-runtime/11.3.0-
r0/dummylib/libstdc++.so: file is empty
collect2: error: ld returned 1 exit status
configure:15686: $? = 1
It complains dummylib/libstdc++.so is empty, configure failed then
wrongly considered S_ISREG doesn't exist.
This seems related to the following commit.
https://github.com/openembedded/openembedded-core/commit/34b0edb0d3120c32063ff7e3dd52be20d60401d5
Reverting this commit rectifies this issue but not sure what's the
proper fix?
Any help appreciated.
There is some piece we're still not understanding. I had a look at my
local build (master so gcc 12.1) and I see HAVE_S_ISREG is set
correctly, I don't see any "file is empty" message.
Are you able to reproduce this on master? Which linker are you using?
G'day Richard,
I've update to master and tried this. Still get a "file is empty" error during
configure. See log snippet below.
Tried to capture the ld version here:
wait4(23043,
/home/preid/dev/v2022.05/tmp-glibc/work/cortexa9t2hf-neon-emit-linux-gnueabi/gcc-runtime/12.1.0-r0/recipe-sysroot-native/usr/bin/arm-emit-linux-gnueabi/../../libexec/arm-emit-linux-gnueabi/gcc/arm-emit-linux-gnueabi/12.1.0/ld:
error:
/home/preid/dev/v2022.05/tmp-glibc/work/cortexa9t2hf-neon-emit-linux-gnueabi/gcc-runtime/12.1.0-r0/dummylib/libstdc++.so:
file is empty
collect2: error: ld returned 1 exit status
/home/preid/dev/v2022.05/tmp-glibc/work/cortexa9t2hf-neon-emit-linux-gnueabi/gcc-runtime/12.1.0-r0/recipe-sysroot-native/usr/bin/arm-emit-linux-gnueabi/../../libexec/arm-emit-linux-gnueabi/gcc/arm-emit-linux-gnueabi/12.1.0/ld
--version
GNU gold (GNU Binutils 2.38.20220516) 1.16
Copyright (C) 2022 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License version 3 or (at your option) a later version.
This program has absolutely no warranty.
Hope that helps.
It does thanks. It confirms you're using the gold linker which allowed
me to reproduce (I was trying with the default bfd). I think this small
change will fix it:
diff --git a/meta/recipes-devtools/gcc/gcc-runtime.inc
b/meta/recipes-devtools/gcc/gcc-runtime.inc
index c39a0caf8a7..7e920df2d34 100644
--- a/meta/recipes-devtools/gcc/gcc-runtime.inc
+++ b/meta/recipes-devtools/gcc/gcc-runtime.inc
@@ -68,7 +68,8 @@ do_configure () {
# libstdc++ isn't built yet so CXX would error not able to find it
which breaks stdc++'s configure
# tests. Create a dummy empty lib for the purposes of configure.
mkdir -p ${WORKDIR}/dummylib
- touch ${WORKDIR}/dummylib/libstdc++.so
+ touch ${WORKDIR}/dummylib/dummylib.c
+ ${CC} ${WORKDIR}/dummylib/dummylib.c -shared -o
${WORKDIR}/dummylib/libstdc++.so
for d in libgcc ${RUNTIMETARGET}; do
echo "Configuring $d"
rm -rf ${B}/${TARGET_SYS}/$d/
G'day Richard,
Thanks, can confirm that this fixes the problem. Our tests all passed.
Will you submit this as a patch?
Affects us on kirkstone, as well as master branch.
--
Regards
Phil Reid
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#168026):
https://lists.openembedded.org/g/openembedded-core/message/168026
Mute This Topic: https://lists.openembedded.org/mt/92099905/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
